Clinical Pharmacogenetics Implementation
Consortium (CPIC) Guideline for
Pharmacogenetics-Guided Warfarin Dosing:
2017 Update
JA Johnson
1
, KE Caudle
2
, L Gong
3
, M Whirl-Carrillo
3
, CM Stein
4
, SA Scott
5
, MT Lee
6
, BF Gage
7
,
SE Kimmel
8,9
, MA Perera
10
, JL Anderson
11
, M Pirmohamed
12
, TE Klein
3
, NA Limdi
13
, LH Cavallari
1
and
M Wadelius
14
This document is an update to the 2011 Clinical
Pharmacogenetics Implementation Consortium (CPIC)
guideline for CYP2C9 and VKORC1 genotypes and warfarin
dosing. Evidence from the published literature is presented for
CYP2C9, VKORC1, CYP4F2, and rs12777823 genotype-guided
warfarin dosing to achieve a target international normalized ratio
of 2–3 when clinical genotype results are available. In addition,
this updated guideline incorporates recommendations for adult
and pediatric patients that are specific to continental ancestry.
Warfarin is a widely used anticoagulant with a narrow therapeu-
tic index and large interpatient variability in the dose required to
achieve target anticoagulation. Common genetic variants in
CYP2C9, VKORC1, CYP4F2, and the CYP2C cluster (e.g.,
rs12777823), plus known nongenetic factors, account for 50%
of warfarin dose variability. This document is an update to the
2011 Clinical Pharmacogenetics Implementation Consortium
(CPIC) guideline for CYP2C9 and VKORC1 genotypes and war-
farin dosing and aims to assist in the interpretation and use of
CYP2C9, VKORC1, CYP4F2, and rs12777823 genotypes to esti-
mate therapeutic warfarin dose among patients with a target
international normalized ratio (INR) of 2–3, should clinical
genotype results be available to the clinician. The CPIC of the
National Institutes of Health’s Pharmacogenomics Research Net-
work develops peer-reviewed gene/drug guidelines that are pub-
lished and updated periodically on https://cpicpgx.org/
guidelines/ and http://www.pharmgkb.org based upon new
developments in the field.
1
These guidelines were written with a
global audience in mind, although the majority of the data that
underpin these guidelines arise from people of European ancestry,
East Asia, and African Americans.
FOCUSED LITERATURE REVIEW
The Supplement includes a systematic literature review on
CYP2C9, VKORC1, CYP4F2 and other relevant genes/ge notypes
that have been associated with warfarin dosing. This systematic
review forms the basis for the recommendations contained in this
guideline. Although some of these genes have also been associated
with dose of other coumarin anticoagulants, the recommendations
below are specific to warfarin.
DRUG: WARFARIN
Warfarin (Coumadin and others) is the most commonly used
oral anticoagulant worldwide, with annual prescriptions in the
1
Department of Pharmacotherapy and Translational Research, College of Pharmacy, and Center for Pharmacogenomics, University of Florida, Gainesville,
Florida, USA;
2
Department of Pharmaceutical Sciences, St. Jude Children’s Research Hospital, Memphis, Tennessee, USA;
3
Department of Biomedical Data
Science, Stanford University, Stanford, California, USA;
4
Division of Clinical Pharmacology Vanderbilt Medical School, Nashville, Tennessee, USA;
5
Department
of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai, New York, New York, USA;
6
Laboratory for International Alliance on Genomic
Research, RIKEN Center for Integrative Medical Sciences, Yokohama, Japan; National Center for Genome Medicine; Institute of Biomedical Sciences,
Academia Sinica, Taipei, Taiwan; Genomic Medicine Institute, Geisinger Health system, Danville, Pennsylvania, USA;
7
Department of Internal Medicine,
Washington University in St. Louis, St. Louis, Missouri, USA;
8
Center for Clinical Epidemiology and Biostatistics, University of Pennsylvania School of Medicine,
Philadelphia, Pennsylvania, USA;
9
Department of Medicine and Department of Biostatistics and Epidemiology, University of Pennsylvania School of Medicine,
Philadelphia, Pennsylvania, USA;
10
Department of Medicine, University of Chicago, Chicago, Illinois, USA;
11
Intermountain Heart Institute, Intermountain
Medical Center, and Department of Internal Medicine (Cardiology), University of Utah School of Medicine, Salt Lake City, Utah, USA;
12
Department of Molecular
and Clinical Pharmacology; The Wolfson Centre for Personalised Medicine; Institute of Translational Medicine, University of Liverpool, Liverpool, UK;
13
Department of Neurology and Epidemiology, University of Alabama at Birmingham, Birmingham, Alabama, USA;
14
Department of Medical Sciences, Clinical
Pharmacology and Science for Life Laboratory, Uppsala University, Uppsala, Sweden. Correspondence: JA Johnson (Johnson@cop.ufl.edu)
Received 12 December 2016; accepted 2 February 2017; advance online publication 15 February 2017. doi:10.1002/cpt.668
CLINICAL PHARMACOLOGY & THERAPEUTICS | VOLUME 102 NUMBER 3 | SEPTEMBER 2017 397
CPIC UPDATE
western world typically equaling 0.5–1.5% of the population.
2
It
is prescribed for the treatment and prevention of thromboembol-
ic disorders.
3
Although highly efficacious, warfarin dosing is
notoriously challenging due to its narrow therapeutic index and
wide interindividual variability in dose requirements, even among
patients with the same target INR.
4
Complications from inap-
propriate warfarin dosing are among the most frequently
reported adverse events to the US Food and Drug Administra-
tion (FDA) and one of the most common reasons for emergency
room visits.
5
Warfarin is usually dosed empirically: an initial dose is pre-
scribed, typically followed by at least weekly measurement of the
INR and subsequent dose adjustment. The initial dose is often
based on population averages (e.g., 4–5 mg/day), but in some set-
tings, it is common to use loading doses during the first few days
of anticoagulation. Irrespective of the method used to initiate
warfarin, stable doses to achieve an INR of 2–3 range from
1–20 mg/day. The iterative process to define the appropriate
dose can take weeks to months and during this period, patients
are at increased risk of over- or under-anticoagulation, and thus
risk of bleeding or thromboembolism.
Warfarin pharmacology and pharmacokinetics
Figure 1 highlights key elements of warfarin pharmacology and
pharmacokinetics. Warfarin inhibits vitamin K epoxide reductase
complex
6
and is administered as a racemic mixture, with S-
warfarin being more potent than R-warfarin.
3
GENES: CYP2C9, VKORC1,ANDCYP4F2
There is substantial candidate gene literature evaluating associa-
tions with warfarin dose requirements, as well as several reported
genomewide association studies (Supplemental Tables S1–S7).
The genes with the strongest literature support, and for which we
make recommendations for use in warfarin dosing, are CYP2C9,
VKORC1, and CYP4F2. Additionally, genomewide association
studies have identified an independently significant single nucleo-
tide polymorphism (SNP) in the CYP2C cluster,
7
which has also
been incorporated into this updated recommendation.
CYP2C9 and warfarin
CYP2C9 is a hepatic drug-metabolizing enzyme in the cyto-
chrome P450 (CYP450) superfamily,
8
and is the primary metab-
olizing enzyme of S-warfarin (Figure 1). CYP2C9 has over 60
known variant alleles (http://www.cypalleles.ki.se/cyp2c9.htm;
CYP2C9 allele definition table
9
). Individuals homozygous for
the reference CYP2C9 allele (CYP2C9*1) have the “normal
metabolizer” phenotype. Each named CYP2C9 star (*) allele is
defined by one or more specific SNPs and, to date, and 18 alleles
have been associated with decreased enzyme activity (CYP2C9
allele definition table
9
). The two most common decreased func-
tion alleles among individuals of European ancestry are
CYP2C9*2 (c.430C>T; p.Arg144Cys; rs1799853) and
CYP2C9*3 (c.1075A>C; p.Ile359Leu; rs1057910).
8
CYP2C9
allele frequencies differ between racial/ethnic groups.
8,10
In vitro and in vivo studies suggest CYP2C9*2 and *3 impair
metabolism of S-warfarin by 30–40% and 80–90%, respec-
tively.
8
Compared to patients homozygous for CYP2C9*1, indi-
viduals who inherit one or two copies of CYP2C9*2 or *3 are at
greater risk of bleeding during warfarin therapy,
11,12
require lower
doses to achieve similar levels of anticoagulation, and require
more time to achieve a stable INR
11
(Supplemental Table S1).
Additional CYP2C9 alleles (CYP2C9*5, *6, *8, and *11) are asso-
ciated with decreased function of the CYP2C9 enzyme and
Figure 1 Schematic representation of warfarin metabolism and its mechanism of action. Warfarin is administered via a racemic mixture of the R- and S-
stereoisomers. S-warfarin is 3–5 times more potent than R-warfarin and is metabolized predominantly to 7- and 6-hydroxyl metabolites via CYP2C9. Warfa-
rin exerts its anticoagulant effect through inhibition of its molecular target VKORC1, which in turn limits availability of reduced vitamin K, leading to
decreased formation of functionally active clotting factors. These clotting factors are glycoproteins that are posttranslationally carboxylated by gamma-
glutamyl carboxylase (GGCX) to Gla-containing proteins. The endoplasmic reticulum chaperone protein calumenin (CALU) can bind to and inhibit GGCX
activity. The metabolism of reduced vitamin K to hydroxyvitamin K1 is catalyzed by CYP4F2, which removes vitamin K from the vitamin K cycle (adapted
from warfarin pharmacokinetics (PK) and pharmacodynamics (PD) pathways at PharmGKB, http://www.pharmgkb.org/do/
serve?objId5PA451906&objCls5Drug#tabview5tab4).
CPIC UPDATE
398 VOLUME 102 NUMBER 3 | SEPTEMBER 2017 | www.cpt-journal.com
contribute to dose variability. These alleles are found with the
highest frequency among those of African ancestry, and collec-
tively are more common than CYP2C9*2 and *3 in that popula-
tion (CYP2C9 frequency table
9
).
VKORC1 and warfarin
VKORC1 encodes the vitamin K epoxide reductase protein, the
target enzyme of warfarin.
6
VKORC1 catalyzes the conversion of
vitamin K-epoxide to vitamin K, which is the rate-limiting step
in vitamin K recycling.
13
A common variant upstream of VKORC1 (c.-1639G> A,
rs9923231) is significantly associated with warfarin sensitivity and
patients with one or two –1639A require progressively lower warfa-
rin doses than –1639G/G homozygotes.
10,14–18
The –1639G>A
polymorphism is present on a haplotype that affects VKORC1 pro-
tein expression (VKORC1 allele definition table
19
).
18
Other common VKORC1 SNPs or haplotypes do not further
improve warfarin dose prediction.
10,16
The c.-1639G>A allele
frequency varies among different ancestral populations
(VKORC1 frequency table
19
), and largely explains the differ-
ences in average dose requirements between whites, blacks, and
Asians.
10,17
Several rare nonsynonymous VKORC1 variants con-
fer warfarin resistance (high dose requirements) and are detailed
in Supplemental Table S2.
20
CYP4F2 and warfarin
CYP4F2 is a primary liver vitamin K oxidase that catalyzes the
metabolism of vitamin K to hydroxy-vitamin K1 and removes
vitamin K from the vitamin K cycle
21
(Figure 1). It acts as an
important counterpart to VKORC1 in limiting excessive accu-
mulation of vitamin K. The nonsynonymous variant CYP4F2*3
(c.1297G>A; p.Val433Met; rs2108622) was first shown to affect
enzyme activity and associated with warfarin dose in three inde-
pendent white cohorts (CYP4F2 gene definition table
24
).
22–24
Furthermore, including this CYP4F2 variant in warfarin dosing
models that included CYP2C9, VKORC1, and clinical factors
improved the accuracy of dose prediction.
25
This correlation has
been confirmed in subsequent studies with those of European
and Asian ancestry, although not those of African ancestry.
26,27
Two large meta-analyses (one in Han Chinese that pulled in sub-
stantial Chinese literature) provide the best estimates for the
influence data of CYP4F2*3 on warfarin dose requirements.
26,27
They suggest statistically significant but modest impacts of 8–
11% higher warfarin doses in A allele carriers (Supplemental
Table S3).
CYP2C rs12777823 and warfarin
rs12777823 is a SNP in the CYP2C cluster near the CYP2C18
gene on chromosome 10 and is associated with a clinically rele-
vant effect on warfarin dose through significant alterations in
warfarin clearance, independent of CYP2C9*2 and *3.
7
This
association was first identified through a genome-wide association
study in African Americans (P 5 1.51310
-8
) and confirmed in a
replication cohort (P 5 5.04310
-5
); meta-analysis of the two
cohorts together produced a P value of 4.5310
-12
. This study
concluded that African Americans who are heterozygous or
homozygous for the rs12777823 A allele require a dose reduction
of 7 or 9 mg/week, respectively.
7
Regression analysis showed
that addition of this SNP improves the dosing algorithm pub-
lished by the International Warfarin Pharmacogenetics Consor-
tium (IWPC) by 21%. Further studies have demonstrated the
importance of this SNP in African Americans.
28
Although this
variant is common in other ethnic populations, an association
with warfarin dose has only been detected among African Ameri-
cans, suggesting it is not the underlying cause but likely inherited
with other variant(s) on a haplotype that influences warfarin
dose in this population. Of note, an association was not observed
in a cohort of Egyptians, thus it is not possible to make broad
statements about this allele in people of continental African
ancestry. Most African Americans are of West African ancestry;
it is unknown whether similar associations are present in individ-
uals from other parts of Africa.
Genetic test interpretation
CYP2C9. Clinical laboratories typically report CYP2C9 genotype
results using the star (*) allele nomenclature system and an inter-
pretation that includes a predicted metabolizer phenotype
(CYP2C9 allele definition table
9
). Most FDA-approved
CYP2C9 tests include only *2 and *3, which is not as informative
for African ancestry populations; however, some clinical laborato-
ries may offer expanded CYP2C9 panels validated as laboratory
developed tests (LDTs) (for allele frequencies see: CYP2C9 fre-
quency table
9
).
VKORC1. Clinical laboratories typically report VKORC1 geno-
type results by c.-1639G>A (or the linked 1173C>T;
rs9934438) genotype (e.g., G/A) and an interpretation on warfa-
rin sensitivity (VKORC1 allele definition table
19
). Most com-
mercial genotyping platforms do not detect rare VKORC1
variants that have been associated with warfarin resistance
(VKORC1 frequency table
19
).
CYP4F2. Although not as commonly tested for as CYP2C9 and
VKORC1, some clinical laboratories may also test for CYP4F2
using a targeted genotyping laboratory developed test to detect
CYP4F2*3 (c.1297G>A, p.Val433Met; rs2108622) variant
(CYP4F2 allele definition table
24
). Results are typically
reported by nucleotide (e.g., G/A), amino acid (e.g., Val/Met) or
star (*) allele (*1/*3) genotype and an interpretation related to
warfarin dosing.
CYP2C rs12777823. Given the recent identification of the associ-
ation between rs12777823 (g.96405502G>A) and warfarin dos-
ing among African Americans, most clinical laboratories do not
currently include this non-coding variant in their warfarin phar-
macogenetic genotyping panels. However, the increasing accessi-
bility of clinical research genomics programs that return
actionable results and the notable effect of this variant among
African Americans suggests that some patients may have geno-
type results for this variant in the future. Results would likely be
reported by genotype (e.g., G/A) and an interpretation related to
warfarin dosing.
CPIC UPDATE
CLINICAL PHARMACOLOGY & THERAPEUTICS | VOLUME 102 NUMBER 3 | SEPTEMBER 2017 399
Genetic test options
Commercially available genetic testing options change over time.
Additional information about pharmacogenetic testing can be
found at the Genetic Testing Registry (http://www.ncbi.nlm.nih.
gov/gtr/).
Incidental findings
No diseases have been linked to common CYP2C9 variants inde-
pendent of drug metabolism and response. Similarly, no diseases
have been consistently linked to common VKORC1 and
CYP4F2 variants that are interrogated in warfarin response tests.
However, homozygosity for rare coding mutations in VKORC1
are a known cause of combined deficiency of vitamin K-
dependent clotting factors-2 (VKCFD2), which is a rare and
potentially fatal bleeding disorder that can be reversed by oral
administration of vitamin K.
29
Linking genetic variability to variability in drug-related
phenotypes
Common variants in CYP2C9, VKORC1, and CYP4F2 account
for up to 18%, 30%, and 11% respectively, of the variance in sta-
ble warfarin dose among patients of European ances-
try,
10,16,17,30,31
but because of differing allele frequencies across
populations, these variants explain less of the dose variability in
patients of other ancestries. In particular, CYP2C9*2 is virtually
absent in Asians, and additional CYP2C9 alleles (e.g., *5, *6, *8,
and *11 alleles) occur almost exclusively in persons of African
ancestry and contribute to dose variability in this population.
Other genes of potential importance are discussed in the Supple-
mental Material.
Published in 2013, the European Pharmacogenetics of Antico-
agulant Therapy (EU-PACT) and Clarification of Optimal Anti-
coagulation through Genetics (COAG) trials examined the
efficacy of genotype-guided warfarin dosing in randomized con-
trolled trials.
32,33
In a homogenous European population, the
EU-PACT trial showed shorter time to stable dose, improved
percent time in therapeutic range, and reduced number of epi-
sodes with an INR >4 using a pharmacogenetic dosing algorithm
compared to standard dosing.
33
The COAG trial was conducted
in an ethnically diverse cohort with 27% of participants of Afri-
can ancestry.
32
Overall, COAG did not show a difference in time
to stable dose, percent time in therapeutic range, reduction in
number of episodes with INR >4or<2, or bleeding risk with a
pharmacogenetic dosing algorithm compared to a clinical algo-
rithm. In nonblacks, the pharmacogenetic dosing algorithm arm
had more patients whose stable dose was within 1 mg per day of
the algorithm-predicted dose (57 vs. 39%, respectively). In con-
trast, the pharmacogenetic dosing algorithm was less accurate at
predicting within 1 mg/day of the stable dose than the clinical
algorithm in blacks (38 vs. 48%, respectively).
32
Blacks were more
likely to have an INR above range with pharmacogenetic dosing,
which could be due to the genotyping panel in the COAG trial
being limited to CYP2C9*2, *3, and VKORC1 c.-1639G>A.
Other variants that influence warfarin dose and are more com-
mon in blacks (i.e., CYP2C9*5, *6, *8, and *11 and rs12777823)
were not genotyped in the COAG trial and their absence likely
led to significant overdosing in patients with these alleles.
10,34
Consequently, this updated CPIC guideline recommends against
pharmacogenetic dosing of warfarin in blacks when only
CYP2C9*2 and *3 genotype results are available.
The Genetics-InFormatics Trial (GIFT) was a randomized
controlled trial examining the effectiveness and safety of geno-
type-guided dosing versus clinical algorithm dosing in orthopedic
patients with a composite outcome that included symptomatic
and asymptomatic venous thromboembolism, major hemorrhage,
INR 4, and death.
35
It is the first warfarin pharmacogen etics
trial powered for clinical outcomes. GIFT included genotyping
for CYP2C9*2 and *3, CYP4F2*3, and VKORC1-1639, but did
not include the African-specific CYP2C9 alleles or rs12777823.
The results of GIFT were presented in early 2017 and revealed a
27% reduction in the composite outcome with genotype-guided
versus clinical algorithm dosing, documenting the clinical benefits
of a genotype guided approach to warfarin dosing. (https://www.
sciencedaily.com/releases/2017/03/170320091104.htm)
Therapeutic Recommendations: Adults
Recommendations for warfarin maintenance (chronic) dosage based
on genetic information.
We use the three-tiered rating system
described previously (and in the Supplemental Material)
1
in
which ratings of strong, moderate, and optional are applied based
on the evidence reviewed. The recommendations for dosing based
on genotype contained herein include recommendations and are
derived from numerous observational and prospective studies, and
randomized trials that suggest the ability to more accurately identi-
fy stable therapeutic warfarin dose requirements through the use of
both genetic and clinical information. Data from prospective stud-
ies and randomized controlled trials are equivocal on whether the
improvement in dosing prediction by pharmacogenetics dosing
leads to improved clinical outcomes. The majority of the literature
underpinning these guidelines arises from individuals of European
ancestry, African Americans, and East Asians. However, the more
limited literature in other populations generally suggests the guide-
lines are appropriate in them also.
Numerous studies have derived warfarin dosing algorithms that
use both genetic and nongenetic factors to predict warfarin
dose.
16,17,36,37
Two algorithms perform well in estimating stable
warfarin dose
16,17
and were created using more than 5,000 subjects,
although as noted above, more recent data suggest they do not per-
form acceptably in African Americans when used without modifi-
cation for CYP2C9 alleles frequently found in the African
population.
32
The Gage and IWPC algorithms or minor adjust-
ments to them have also been the algorithms used in both random-
ized controlled trials and most of the prospective dosing studies.
Dosing algorithms using genetic information outperform nongenet-
ic clinical algorithms and fixed-dose approaches in dose prediction,
except in African Americans when the algorithm only includes
CYP2C9*2 and *3.
16,17,32
Genetics-based algorithms also better pre-
dict warfarin dose than the FDA-approved warfarin label table.
38
Pharmacogenetic algorithm-based warfarin dosing. This guideline
recommends that pharmacogenetic warfarin dosing be accom-
plished through the use of one of the pharmacogenetic dosing
CPIC UPDATE
400 VOLUME 102 NUMBER 3 | SEPTEMBER 2017 | www.cpt-journal.com
algorithms described above, as summarized in Figure 2. These
algorithms, as originally published, are available in the Supple-
ment and the dosing algorithm published by IWPC is also online
at http://www.pharmgkb.org/do/serve?objId5PA162372936&
objCls5Dataset#tabview5tab2. The two algorithms provide very
similar dose recommendati ons. The clinical and genetic information
used in one or both algorithms is shown in Text Box 1.Thesealgo-
rithms compute the anticipated stable daily warfarin dose to one
decimal and the clinician must then prescribe a regimen (e.g., an esti-
mate of 4.3 mg/day might be given as 4 mg daily except 5 mg 2 days
per week). An additional “dose revision” algorithm, which can be
used on days 4–5 of therapy for dose refinement and uses genet ic
information, was tested in COAG and EU-PACT and can also be
used
36
(Su pplement al Table S5).
It is important to note that these algorithms do not include
CYP4F2, CYP2C9*5, *6, *8,or*11 or rs12777823, and incorpo-
ration of these should be added when results are available, as
described in Figure 2. The warfarindosing.org website contains
both algorithms, the Gage algorithm
16
as the primary algorithm
and the IWPC algorithm
17
as the secondary algorithm and
can adjust for CYP4F2, CYP2C9*5, and *6. If utilizing
warfarindosing.org, the user should be clear on whether the
Text Box 1. Patient characteristics utilized in the Gage (16), or IWPC (17)
algorithms or both
Age
Sex
Race
Weight
Height
Smoking status
Warfarin indication
Target INR
Interacting drugs
䊊 Inhibitors: Amiodarone, statins, sulfamethoxazole, azole
antifungals
䊊 Inducers: Rifampin, phenytoin, carbamazepine
Genetic variables
䊊 CYP2C9 genotype
䊊 VKORC1 genotype
䊊 Gage algorithm can also incorporate CYP4F2 and GGCX
genotypes
Figure 2 Dosing recommendations for warfarin dosing based on genotype for adult patients. (a) “Dose clinically” means to dose without genetic informa-
tion, which may include use of a clinical dosing algorithm or standard dose approach. (b) Data strongest for European and East Asian ancestry popula-
tions and consistent in other populations. (c) 45–50% of individuals with self-reported African ancestry carry CYP2C9*5, *6, *8, *11, or rs12777823. If
CYP2C9*5, *6, *8, and *11 were not tested, dose warfarin clinically. Note: these data derive primarily from African Americans, who are largely from West
Africa. It is unknown if the same associations are present for those from other parts of Africa. (d) Most algorithms are developed for the target INR 2-3.
(e) Consider an alternative agent in individuals with genotypes associated with CYP2C9 poor metabolism (e.g., CYP2C9*3/*3, *2/*3, *3/*3) or both
increased sensitivity (VKORC1 A/G or A/A) and CYP2C9 poor metabolism. (f) See the EU-PACT trial for pharmacogenetics-based warfarin initiation (load-
ing) dose algorithm with the caveat that the loading dose algorithm has not been specifically tested or validated in populations of African ancestry.
33
(g) Larger dose reduction might be needed in variant homozygotes (i.e., 20–40%). (h) African American refers to individuals mainly originating from West
Africa.
CPIC UPDATE
CLINICAL PHARMACOLOGY & THERAPEUTICS | VOLUME 102 NUMBER 3 | SEPTEMBER 2017 401